A clinical approach for the diagnosis of diabetes mellitus: an analysis using glycosylated hemoglobin levels. Meta-analysis Research Group on the Diagnosis of Diabetes Using Glycated Hemoglobin Levels

Abstract

To determine whether a glycosylated hemoglobin level can be used in place of an oral glucose tolerance test (OGTT) to diagnose diabetes. An augmented MEDLINE search was performed to identify all reports from 1966 through June 1994 in which glycosylated hemoglobin levels were measured concurrently with performance of OGTTs in the same study. The corresponding authors were contacted and asked to provide individual data for all subjects tested. A total of 31 investigators representing 34 possible studies responded, and 18 were able to provide us with the data requested. Overall fasting plasma glucose concentrations, 2-hour postdextrose glucose concentrations, and glycosylated hemoglobin levels were available from 11 276 individuals. To define normal glucose tolerance, impaired glucose tolerance (IGT), and diabetes, modified World Health Organization criteria were used. An analysis of the methods used for measurement of glycosylated hemoglobin levels revealed that the HbA1c assay showed the least variance in normal subjects. Therefore, only data from the 8984 subjects who had HbA1c levels measured were used. When we used the mean HbA1c level plus 4 SDs as a cutpoint, the sensitivity was 36% and specificity was 100% compared with the results of the OGTT. Because of the lack of agreement between OGTT results and HbA1c levels, models were created to analyze the distribution of HbA1c levels in each study. Using these models, we identified 3 subpopulations. The third subpopulation was likely to represent subjects with diabetes. When we applied an HbA1c level of 7.0% as a cutpoint, the sensitivity was 99.6% for the third subpopulation. When this cutpoint was reapplied to the OGTT results, of those subjects with an HbA1c level of at least 7.0%, 89% had diabetes, 7% had IGT, and 4% were normal. Although the OGTT is the "gold standard" for diagnosing diabetes, it is known to be poorly reproducible and is often not performed. Not only is use of an HbA1c level to diagnose diabetes more convenient, but therapeutic decisions are based on this value, regardless of the findings on the OGTT. An HbA1c level of 7.0% or higher often requires pharmacological intervention and is most often associated with the diagnosis of diabetes by World Health Organization standards. An HbA1c level below 7.0% would generally be treated with diet and exercise, regardless of the diagnosis of IGT or diabetes by OGTT. Thus, measurement of HbA1c levels may represent a reasonable approach to identifying treatment-requiring diabetes.